Re: [Qemu-devel] Re: [PATCH master+0.14 0/2] blockdev memory leaks

[Anthony Liguori]

On 02/09/2011 08:09 AM, Justin M. Forbes wrote:
> On Wed, 2011-02-09 at 14:52 +0100, Kevin Wolf wrote:
>    
> > Hi Justin,
> >
> > Am 08.02.2011 15:12, schrieb Markus Armbruster:
> >      
> > > Markus Armbruster (2):
> > >    blockdev: Plug memory leak in drive_uninit()
> > >    blockdev: Plug memory leak in drive_init() error paths
> > >
> > >   blockdev.c |   12 ++++++++++--
> > >   1 files changed, 10 insertions(+), 2 deletions(-)
> > >        
> > How this series made its way into stable was a bit surprising for me.
> > You may not be aware yet of the expectations that I (and probably
> > others) have in the process of patches being applied to stable. No harm
> > done, but maybe something to consider for future patches, so let me just
> > mention some points:
> >
> > I saw that you already merged these patches into the stable tree, even
> > though they are not master yet. I think usually stable should only get
> > cherry-picks from master. There are exceptions of course (e.g. when
> > something will be fixed differently in master), but I don't think this
> > is one of them.
> >
> > Also I noticed that you didn't add your Signed-off-by when applying the
> > patches. As I understand it, you should do this for any patch that you
> > apply directly (i.e. that you don't get via a git pull)
> >
> > I only caught this by chance. If you sent an email ("Thanks, applied to
> > ...") after you have applied a patch or pulled from somewhere, it would
> > be more obvious to the rest of us what happens in stable.
> >      
> Indeed, that was my fault...

No worries, it happens.

Regards,

Anthony Liguori

>   I had applied them for testing, and pushed
> to the wrong tree.  I have made some local changes to insure that this
> does not happen in the future.  The process is, pathches should be in
> master first, or their be a very good reason that they are not (bugs no
> longer apply to master, or different and more invasive fix is used in
> master).  At this point in the cycle, where there is so little
> divergence from master, there is no reason that anything in stable is
> not in master.
>
> Justin
>
>
>